Machine Learning (ML)
Deep Learning
Natural Language Processing (NLP)
Embeddings
Network Science/Analysis
Computer Vision
Swarm Flight
I am Gérome Ferrand, a machine learning engineer from France with a background in computer science. I am actively seeking and participating in diverse projects where machine learning can offer solutions.
Information processing cannot exist without sharing information upstream.
This website hosts various content related to my personal researches, ongoing projects, application cases, and other details about my work. Visitors can access code that covers the complete data lifecycle.
In addition, students will have access to diverse resources aimed at enhancing their knowledge and skills in the areas of data science, data pattern recognition, and machine learning model development.
Every data project requires data and therefore starts with data acquisition. The method of acquisition varies depending on the project: scraping, API request and sensor management are examples.
The acquired data must be cleaned, formatted and finally stored. Depending on the characteristics of the collected data, SQL or noSQL databases are required for this type of task.
Data mining allows us to discover exploitable patterns in order to transform them into added value for our project. A strong background in mathematics may be necessary for this step.
Although it can be very easy to create visual representations, it is much more difficult to create good ones. Tableau is a tool that allows to make excellent adapted visualisations.
Once an initial analysis has been carried out and a proof of concept satisfied, it may be interesting to automate decisions and more generally the transfer of knowledge from new cases. Numerous libraries allow to achieve this (SKLearn, PyTorch...).
A rocket is nothing without its launcher. When you want to deploy a machine learning model in production, it is necessary to use technologies such as AWS or Microsoft Azure to maintain it.
You will find here the latest posts published in the Blog section.
Cette dernière partie aborde les alternatives optimales de ReLU pour améliorer vos ANN, Ainsi que 2 autres fonctions plus spécifiques mais tout autant performantes.
deep learning, basesCette partie explore les différentes fonctions de type rectifier, issue de ReLU. Ces fonctions sont les plus utilisée dans le domaine du Deep Learning.
deep learning, basesDans cette partie: Explorez les avantages et limites des fonction d'activations basiques, ainsi que leur lien avec les neurones biologiques.
deep learning, bases